提出抗干扰的去中心化多臂赌博机算法，有效应对恶意攻击和故障节点。
Robust Decentralized Multi-armed Bandits: From Corruption-Resilience to Byzantine-Resilience
- 设计新算法DeMABAR，用可扩展机制抵御奖励数据污染。
- 在有限攻击预算下，个体累积损失仅额外增加与攻击量成比例项。
- 天然适用于拜占庭场景，适合高可靠性分布式系统应用。
去中心化协同多智能体多臂赌博机（DeCMA2B）研究多个智能体在去中心化环境下的协作策略。尽管已有大量研究，但现有方法对各类对抗性攻击仍敏感。本文首先研究存在对抗性污染的DeCMA2B问题，其中敌手可在有限污染预算内篡改所有智能体的奖励观测。我们提出稳健算法DeMABAR，保证每个智能体的个体遗憾仅增加与污染预算成比例的附加项。随后考虑更现实的情形：敌手仅能攻击少数智能体。理论分析表明，该算法几乎完全消除恶意攻击影响，具备内在拜占庭鲁棒性——即未知比例的智能体可能为拜占庭节点，任意选择动作并传播错误信息。数值实验验证了该方法的鲁棒性与有效性。
原文摘要 · Abstract (English)
Decentralized cooperative multi-agent multi-armed bandits (DeCMA2B) considers how multiple agents collaborate in a decentralized multi-armed bandit setting. Though this problem has been extensively studied in previous work, most existing methods remain susceptible to various adversarial attacks. In this paper, we first study DeCMA2B with adversarial corruption, where an adversary can corrupt reward observations of all agents with a limited corruption budget. We propose a robust algorithm, called DeMABAR, which ensures that each agent's individual regret suffers only an additive term proportional to the corruption budget. Then we consider a more realistic scenario where the adversary can only attack a small number of agents. Our theoretical analysis shows that the DeMABAR algorithm can also almost completely eliminate the influence of adversarial attacks and is inherently robust in the Byzantine setting, where an unknown fraction of the agents can be Byzantine, i.e., may arbitrarily select arms and communicate wrong information. We also conduct numerical experiments to illustrate the robustness and effectiveness of the proposed method.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。